XPost: alt.astronomy   
      
   As I've indicated a few times on these groups, data-science is starting   
   to make gains in determining chains of causation.  In much the same   
   way that a criminal leaves something of themselves at a crime scene,   
   and the crime scene also leaves something on the criminal, one dataset   
   that represents an "input" to a process will nominally leave traces on   
   any dataset observed as "output" from the process.   
      
   In the work I'm doing on looking at UFO data we've seen e.g. the   
   "fingerprint" of the changing distance between certain planets and the   
   Earth seems to appear inside the day-by-day UFO sightings data.  This   
   is evidence that those planets beyond just a simple chance change the   
   distribution of UFO sightings "for some reason".  We might even   
   suspect some of the objects "come" from those planets although another   
   intriguing possibility is they are "radio controlled" from there.   
      
   The best statistical models I have at the moment show that certain   
   time-shifted datasets relating to the day-by-day distance between the   
   Earth and some other planet closely predict UFO sightings. More   
   closely than could happen by chance alone.   
      
   The "time shift" or lag points to the intriguing possibility we could   
   estimate the time it takes for these objects to travel between the   
   Earth and other planets.   
      
   This possibility is further enforce by simple simulations that try to   
   follow the track of probabilistic UFO's travelling between various   
   planets in the solar system. By running through many simple strategies   
   we can find some that also closely match the observed day by day   
   distribution of UFO sightings.   
      
   By no real coincidence the estimated speed from the "lagging method"   
   closely matches the assumed-constant straight-line speed in the   
   simple simulations.   
      
   This is all pointing at an "average" or "drift" speed of the objects   
   rather than actually measuring interplanetary flight times.  Even 2   
   different methods getting the same results within narrow error bounds   
   only shows a naive model of what may be going on.  Interestingly, the   
   speed both these things comes up with is around solar escape speed and   
   nowhere near interstellar speeds unless you are very long-lived. :)   
      
   But the set of results DOES include what normally you might think of   
   as "noisy" ones that might point to interstellar travel.   
      
   So that's where the next phase of the research is landing. Try to   
   find a dataset associated with various stars that we might try to   
   search for in the UFO sightings data. Is there a star -- presumably a   
   nearby star -- where something it does that can be measured from Earth   
   leaves a "fingerprint" in the daily UFO sightings over, say, the last   
   70 years?   
      
   To give you a hint where I'm going here's a model the simple AI s/w   
   I'm using found that links the "light curve" of Uranus delayed by a   
   couple months with cleaned-up UFO sighting data.   
      
   Year/month	   Avg Ur mag	Avg #UFO seen  Model prediction for #UFOs   
   1970.38             5.35589      302.256      335.466   
   1975.46              5.3916      420.969      348.303   
   1975.54             5.40398      488.279      352.754*(model sd under obs)   
   1977.54             5.41776      611.303      357.707**(-2s)   
   1978.54             5.43493      412.838       363.88   
   1979.54             5.44943      430.238      369.092   
   1980.54             5.46202      471.263      373.618   
   1981.54              5.4761      404.493       378.68   
   1982.62             5.49327      273.717      384.852*(+1)   
   1955.04             5.50471      274.724      388.965*(+1)   
   1983.62             5.51195      335.456      391.567   
   1985.62             5.53189      332.517      398.736   
   1986.62             5.55212      237.961      406.008*(+1)   
   1987.62             5.56784      351.138      411.659   
   1988.71             5.58764      332.272      418.777   
   1989.71             5.60454      292.209      424.852*(+1)   
   1990.71             5.62213      280.397      431.176*(+1)   
   1991.71             5.63359      471.281      435.295   
   1992.71             5.64636      366.984      439.886   
   1994.71             5.66503      313.716      446.598*   
   1995.71             5.68515      393.069       453.83   
   2021.04             5.70423      401.085      460.689   
   2020.04             5.72355       449.71      467.635   
   2021.12             5.73537      449.526      471.884   
   2020.12             5.75032      586.626      477.258*   
   2013.96             5.76173      648.603       481.36*   
   2019.12             5.77073       615.72      484.595*   
   2021.21             5.79195      614.015      492.224*   
   1988.12             5.79954      414.908      494.952   
   2020.21             5.80827      617.884       498.09*   
   2019.21             5.82495      478.952      504.087   
   2021.29             5.84342      681.031      510.726*   
   2020.29             5.85754      666.414      515.802*   
   2021.62             5.87652      522.891      522.625   
   2020.62               5.897      648.566      529.988*   
   2021.54             5.91288      529.564      535.696   
   2019.54               5.929      590.398      541.491   
   2018.46             5.94564      666.516      547.473*   
   2016.46             5.95699      470.315      551.553   
   2015.46              5.9703      420.087      556.338*   
   2011.38             5.98229      446.148      560.648*   
      
   MODEL:   
   (AUTO CORR CORRECTION; estimated rho = 0.269625)   
   y = 359.487*x + -1589.9   
   beta in 359.487 +- 179.494  90% CI   
   alpha in -1589.9 +- 746.898   
   T-test: P(beta>0) = 0.999177   
   Rank test: calculated Spearman corr = 0.504681   
   	Critical value = 0.432 2-sided at 1%; reject H0:not_connected   
   r2 = 0.21695668   
      
   I.e. the fingerprint of the mag for Uranus delayed by 2m is found in   
   the UFO sightings numbers so much so that the simple formula follows   
   the ups and down of about 22% of UFO sightings.  Pretty interesting   
   given 90% of UFO sightings are often claimed to be mundane objects   
   that have nothing to do with interplanetary things.   
      
   The big "drawback" with this procedure to date is the data it's using   
   is a "calculated: magnitude for the planets produced by a simple   
   program.  I inherited the program from some guys at an amateur   
   astronomy equipment store I used to live at in the 1980s.  They had   
   tinkered up something to aim their home-made equipment -- some of it   
   of course pretty impressive -- and we managed to translate it from   
   FORTRAN-IV to gee-whiz C and update the formulas to epoch 2000.  It   
   seemed to work OK for calculating RA and Dec but I have since learned   
   the way it performed the calculations from models for the orbits of   
   the planets may predict RA and Dec to a fraction of a degree over a   
   50y period, but the position estimates for the planets is way way out.   
   E.g. the perihelion the program predicts for Jupiter are typically a   
   week out and for outer planets can be a month out.   
      
   So the next step up will be to use actual observations to see whether   
   *observed* positions and magnitude leave an even clearer imprint in   
   the UFO data than the model for the quantities does.   
      
   Given that works out even approximately, the next step is to find   
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)